Comorbidities Among Mental Disorders in Preschool Children in the Mercy Pregnancy Emotional Wellbeing Study

Study Design

Data from the Mercy Pregnancy Emotional Wellbeing Study cohort are used for this study. Further details about this study can be found in the cohort profile paper (Galbally et al., 2017). Women completed Wave 1 data collection at recruitment at up to 20 weeks’ gestation; they were subsequently followed up at up to 28 weeks’ gestation (Wave 2); subsequent data about both the mothers and their children were collected at six months postpartum (Wave 4), 12 months postpartum (Wave 5) and when their children were 3 to 4 years old (Wave 6). At birth, information was captured from hospital records but not from women themselves (Wave 3). Data from Wave 6 (child age 3–4 years) are included in this paper. Sample characteristics are drawn from Wave 1 (<20 weeks’ gestation) and Wave 3 (birth) data.

Recruitment

As described (Galbally et al., 2017), recruitment took place at metropolitan maternity hospitals in Melbourne, Victoria and Perth, Western Australia, as well as regional and rural hospitals in Western Australia. This was a selected cohort design, in which women formally diagnosed with current depression using the Structured Clinical Interview for DSM-IV (SCID-IV) (Carter et al., 2010) were deliberately over-recruited, as well as a group of non-depressed pregnant women.

Women were eligible to participate if they were less than 20 weeks’ gestation and proficient in English. Exclusion criteria were psychotic or substance abuse disorders, child protection involvement, intellectual disability, serious physical illness or psychiatric illness requiring current acute inpatient admission.

Measures

Child mental disorder: The primary outcome was a diagnostic classification of one or more mental disorder among 3–4-year-old children according to DSM-5 criteria (American Psychiatric Association, 2013), and assessed by the Preschool Age Psychiatric Assessment (PAPA) (Egger et al., 1999). The PAPA is a semi-structured research diagnostic instrument that measures psychiatric symptoms and disorders according to the DSM-5, for pre-school age children, aged 2-5. The e-measures electronic version of the PAPA was administered through interviews with mothers of 3-4 years old within MPEWS. The research team undertook training with the authorised trainers of this instrument from Duke University, Developmental Epidemiology (Egger & Angold, 2006), which included coding practice and checking coding with trainers. Following the initial training, subsequent interview administrators underwent a month of training with experienced interviewers who had been trained in the original training with PAPA developers. To facilitate inter-rater reliability, the novice interviewers also completed the coding of at least three interviews after training whilst under supervision with experienced interviewers before completing coding on their own. To ensure quality control, protocol fidelity and consistency for the duration of the data collection, regular supervision with experienced trainers was offered fortnightly to administrators, in which coding queries were discussed with other novice and experienced interviewers.

In the PAPA, compulsory questions are asked about whether symptoms have been present or absent in the past 3 months. If an item is endorsed, the parent is asked follow-up questions, such as onset, frequency, duration and severity of symptoms, and impairment, and to give examples of the behaviour. These behavioural examples provide critical information which guide the interviewer to formally code the PAPA in e-measures. Subsequently an algorithm is applied to determine diagnosis based on coded data. The PAPA has been validated as a reliable standardized measure of DSM-4 psychiatric symptoms and disorders in pre-schoolers (Egger et al., 2006). Its test-retest reliability is comparable to widely used and well-established measures for older children and adults (Egger et al., 2006).

For the purposes of this study, we include the following DSM-5 diagnoses measured using the PAPA: anxiety disorders (separation anxiety, social anxiety, specific phobia, agoraphobia, panic attacks and generalised anxiety disorder (GAD)); post-traumatic stress disorder (PTSD); depressive disorders (including major depression, dysthymia, and/or other specified depressive disorders); attention-deficit/hyperactivity disorder (ADHD), conduct disorder and oppositional defiant disorder (ODD).

As women with a current depressive episode within the past two years were over-represented in this study, we report on the number and proportion of women who met these criteria, as assessed using Modules A and D of the SCID-IV-TR, Clinician Version (First et al., 1997).

Statistical Methods

This is a descriptive study, presenting data on prevalence and patterns of mental disorders and comorbidities in our sample of 3–4-year-old children. Data were analysed using IBM SPSS Statistics version 29.

To evaluate attrition bias, data on sociodemographic characteristics and depression status at <20 weeks’ gestation were compared between women who completed the PAPA (at child age 3–4 years) with women who did not. For binary variables such as depression status, Χ² tests were used, applying the continuity correction. For maternal age, which was not normally distributed, a Mann-Whitney U test was applied.

We report the proportion of children who met DSM-5 criteria for each disorder, and for two or more disorders, according to the PAPA. As a sensitivity analysis we compare the prevalence of two or more mental disorders among children of women who had a current depressive episode at < 20 weeks, and children of women who did not.

Patterns of comorbid mental disorders were first explored by examining the proportion of children diagnosed with each mental disorder, and then their comorbidity patterns with another disorder. For example, we report the proportion of children with a diagnosis of ADHD who also had an additional diagnosis of each other disorder. We demonstrate the degree of comorbidity using a heatmap, generated within Excel using an algorithm to group the data into 6 colour categories; green shades represent lower proportions while red shades indicate higher proportions.

Secondly, we report the odds ratios (OR) and confidence intervals associated with having a specific mental disorder, for each other disorder. Owing to multiple comparisons (approximately 70) and the associated risk of Type I errors, we apply a Bonferroni correction and report only odds ratios with statistical significance <0.001. The only other PAPA-based studies to calculate odds ratios for combinations of mental disorders among preschool children (Bufferd et al., 2011; Wichstrøm et al., 2012) report substantial odds ratios of 5 and above. To detect a more modest odds ratio (OR) of about 2.0 with 80-90% power at a 5% significance level, one requires approximately 16 cases and 16 controls (Lemeshow et al., 1988). Thus, we did not calculate odds ratios within any groups for which exposure (in this case the other mental disorder) did not occur in at least 16 cases.

Ethics

The study has been approved by the Mercy Health Research Ethics Committee (Victoria, R08/22, 3 July 2012) and the South Metropolitan Health Service Human Research Ethics Committee (Western Australia, 2016-192, 11 October 2016). Each woman provided written consent to participate at each wave of data collection. If PAPA interview participants became distressed, the interview was suspended or paused, and the interviewer provided options to abort or reschedule the interview. Information on available supports for women was provided on the Participant Information and Consent form at each wave of data collection, and interviewers discussed these with women as appropriate. Women were also offered a telephone call with a psychiatrist from the research team if they would like to discuss support options. Interviewers offered all women a letter to their child’s general practitioner, psychologist, paediatrician or psychiatrist summarising the results of the research interview. If women indicated they would like this to occur, we obtained written consent from them along with the practitioner’s mailing address.

Strengths and Limitations

This study adds to a small group of studies (Bufferd et al., 2011; Wichstrøm et al., 2012) reporting prevalence of mental disorders and comorbidities among preschool-aged children, using the PAPA structured interviews with parents (Egger et al., 2006, 2019).

We acknowledge several limitations. The PAPA, while widely used and demonstrating high test-retest reliability among preschool children, was adapted from the parent version of the Child and Adolescent Psychiatric Assessment (CAPA) (Angold et al., 1995); other assessments are available that were developed specifically for this age group (In-Albon et al., 2020). Ours was a selected cohort study, not recruited for the purpose of reporting prevalence and hence without the specific methodological requirements for a prevalence study. Mothers with depression diagnosed early in pregnancy are over-represented due to the recruitment strategy and therefore this is not a community sample where generalisable prevalence data can be ascertained. Overreporting of children’s psychiatric symptoms has been reported among depressed mothers (Müller et al., 2011, 2014). We did not assess maternal depression with structured interviews at child age 3–4. Although we took steps to improve statistical power (excluding small groups from odds ratio calculations) and corrected for multiple comparisons, we advise readers to interpret Table 4 with caution owing to some very wide confidence intervals. Given the deliberate over-sampling of women with depression in our longitudinal cohort, the relatively high level of tertiary education, and the fact that only 62% of women recruited during pregnancy completed the PAPA when their children were 3–4 years old, we recommend that this finding be confirmed with a large, more representative sample, controlling for relevant pregnancy and early childhood environmental factors.

Further limitations may include that we administered the PAPA via telephone interview and not face to face as originally recommended by the authors of the PAPA (Egger & Angold, 2004). Information bias leading to false positives may arise from telephone interviewing. Although other studies report telephone interviews for the PAPA (Bufferd et al., 2011; Danzig et al., 2013; Hattangadi et al., 2024), reliability for telephone-based administration is lacking. Owing to limited resources, we were unable to implement double coding of the PAPA interviews beyond the initial training period of novice interviewers; inter-rater reliability can therefore not be reported Our study was a cross-sectional assessment of the children; future research to report on the trajectory and stability of disorders and comorbidities is recommended. It is not known how impairment associated with multiple diagnoses compares with that associated with single diagnoses; a nuanced assessment of impairment for both groups of children could address this in future research. In addition, we collected data only from mothers; it would be preferable to have multiple sources of data and future studies should collect data from fathers or other informants, to reduce informant bias (Bufferd et al., 2011).

The PAPA is a reliable diagnostic instrument for the preschool population. The high prevalence ascertained in our study may arise due to the robustness of this instrument compared with other studies which utilised parent-reported rating scales. However, the high prevalence may also be the result of over-pathologisation of normative responses, such as fear, which is a normative developmental phenomenon among preschool-aged children. As mentioned, the PAPA was adapted from the CAPA; diagnostic criteria for older children have therefore been applied to a younger age group. In addition, we have noted methodological limitations including non-representative sampling, informant bias and information bias arising from telephone interviewing. Our findings should therefore be treated as preliminary and with caution, awaiting replication by other studies in Australia or similar settings that utilise research-based semi-structured diagnostic interviews such as the PAPA. Longitudinal studies would illuminate homo- and heterotypic trajectories in children in non-clinical samples; among a clinical sample of children from preschool to primary school age, heterotypic development was relatively common, especially in the presence of maternal mental disorders (Jobs et al., 2019).

Alternatively, rather than the result of measurement or cohort effects, the high prevalence could be attributable to the preschool developmental window – a developmental vulnerability window which needs to be managed with care. As such, a “preschool-limited vulnerability” contrasting with “life-course-persistent variants” can be hypothesized as analogous to the “adolescence-limited” subtype of antisocial behaviours reported by Moffitt (1993). As noted above, the features of specific phobias, which appear to be particularly prevalent during the preschool period, could be conflated with normal developmental fears, or represent features of a normal developmental phase that will resolve with maturation. Longitudinal follow-up studies can track the developmental trajectories and differentiate whether “preschool-limited” and “life-course-persistent” subtypes exist.

Implications

Preliminary findings from out study, to be interpreted with caution, suggest that among preschool-aged children assessed with the PAPA, almost half yielded at least one diagnosis of a mental disorder. If problems are identified early, adjustments can be made before and once formal education begins, potentially preventing educational impacts from chronicity of mental disorders in later childhood and adulthood (Lyons-Ruth et al., 2017; Weisz et al., 2017). If mental health problems are left undetected, relationships with caregivers and peers may worsen over time and become more resistant to change (Bricker et al., 2004).

The relatively common occurrence of comorbidities (21% of children in our sample, 44% of those with a diagnosis according to the PAPA) suggests that children presenting with any mental disorder should be comprehensively assessed using a multi-methods assessment framework for a broad range of mental disorders. In addition, comorbidities may not be limited to “internalising” and “externalising” categories. When comorbid mental disorders are identified, coordinated child mental health services and treatments need to be accessible for a child and their family (Fisher, 2022). Comorbidity complicates treatment of children (Weisz et al., 2017); treatments that focus on single disorders may not be appropriate and may need to be replaced by treatments that consider multiple disorders (Barlow & Carl, 2012); the evidence for efficacy of such treatments is still limited and mostly among older children (Weisz et al., 2017).

Our findings also raise the possibility whether there exists a “preschool-limited” variant of psychopathology. Such variants can explain the high prevalence, analogous to the steep increase in anti-social behaviours in the adolescence window (Moffitt, 1993). Future studies with longitudinal design can explore this hypothesis; and if confirmed, examine the predictors of “preschool-limited” and “life-course-persistent” subtypes.

Our study suggests that clinicians and researchers should be aware of the relatively high prevalence of mental disorders and various comorbidities among these. Prior to the commencement of formal education, there is a window of opportunity to identify and manage disorders and lay the foundation for more optimal mental health and relationships.